We bought the bad version first.

We bought leads for about a year. We spent roughly fifty thousand dollars doing it. We closed nothing.

For most of that year we believed it was working. The reports came back every week with open rates that looked healthy and a click count that looked reasonable, and we renewed on the strength of them more than once.

When we finally sat down and went through the file properly, the reasons were dull. A large share of the addresses were dead. Plenty of the people had left the company, some of them years earlier. A number of the job titles had been assigned by whatever tool assembled the list rather than by the company the person actually worked at.

None of that was visible on the report, because the report showed opens. Opens look healthy when scanners open everything.

What that year actually taught us

The useful part was not learning that some vendors sell bad data. Everybody knows that. It was learning exactly how a bad file passes inspection: which numbers get reported because they always look good, which ones get left out because they do not, and how long you can run on the difference before anyone opens the spreadsheet.

So when we say we report the smaller number, it is not a philosophy. We know precisely which larger number we would be quoting instead, and we know what it would be hiding, because we spent a year on the other side of that invoice.

Every rule came from that file

Opens are not intent, because opens were the number that kept us renewing. Distinct human clicks, because the raw ones were scanners. No data means no line, because invented titles are what made the list look full. Suppressed contacts stay suppressed, because the same dead names kept coming back on the next import.

It is also why we audit a pipeline before selling into it. The year that cost us fifty thousand dollars did not have a lead volume problem. It had a nobody-checked-the-file problem, and more volume would have made it worse.